Advertisement putting effect analysis method and device and storage medium

Abstract

The invention relates to the technical field of advertisement putting effect analysis methods, in particular to an advertisement putting effect analysis method and device and a storage medium, and the method specifically comprises the following steps: 1, data collection and preprocessing: determining the types of data needing to be collected, including advertisement putting data, user behavior data and market environment data; the method comprises the following steps: setting a data collection period, and determining data collection by taking day, week, month and the like as units according to the duration and characteristics of advertisement putting to ensure the continuity and timeliness of the data: preprocessing the collected data; and 2, constructing a key index system, thereby constructing an index system from four dimensions of a coverage effect, an interaction effect, a conversion effect and a cost effect of advertisement putting. According to the method, the limitation of a single index is broken through, and one-sided cognition is avoided through multi-index linkage analysis and association.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of advertisement delivery effect analysis methods, in particular to an advertisement delivery effect analysis method, device and storage medium. BACKGROUND

[0002] In the digital age, advertisement delivery has become an important means for enterprises to promote products and enhance brand awareness. With the rapid development of Internet technology, advertisement delivery channels are increasingly diversified, such as social media platforms, search engines, e-commerce websites, etc. However, in the face of massive advertisement data and complex delivery environment, how to accurately and comprehensively analyze the effect of advertisement delivery has become a problem that enterprises need to solve.

[0003] In the prior art, advertisement delivery effect analysis methods often have many limitations. Some methods only focus on a single effect indicator, such as click-through rate, conversion rate, etc., ignoring the relevance between different indicators and the dynamic changes in the advertisement delivery process. SUMMARY

[0004] In view of the technical problem that some methods in the art only focus on a single effect indicator, such as click-through rate, conversion rate, etc., ignoring the relevance between different indicators and the dynamic changes in the advertisement delivery process, the present application provides an advertisement delivery effect analysis method, device and storage medium.

[0005] The technical solution adopted by the present application is: an advertisement delivery effect analysis method, specifically comprising the following methods:

[0006] Step one: data collection and preprocessing, clearly defining the types of data to be collected, including advertisement delivery data, user behavior data and market environment data; setting the data collection period, determining to collect data in units of days, weeks, months, etc. according to the duration and characteristics of advertisement delivery, ensuring the continuity and timeliness of the data; preprocessing the collected data;

[0007] Step two: building a key indicator system, thereby building an indicator system from four dimensions of coverage effect, interaction effect, conversion effect and cost effect of advertisement delivery;

[0008] Step three: determining the indicator weight, using the analytic hierarchy process to determine the indicator weight;

[0009] Step four: dynamic effect evaluation model construction;

[0010] Step five: impact factor analysis;

[0011] Step six: effect prediction and optimization suggestion;

[0012] Step 7: Performance tracking and feedback adjustment. Track the performance of the advertising campaign and adjust the analysis methods and advertising strategies in a timely manner based on the feedback results.

[0013] In one embodiment, step one includes data preprocessing, which includes data cleaning and data standardization.

[0014] The data cleaning methods are as follows:

[0015] Remove duplicate data: Identify and delete duplicate records by comparing data. Let the number of duplicate records be D. r The total amount of original data is D t The percentage of duplicate data

[0016] When R r If the threshold is exceeded, it is necessary to check for problems in the data collection process;

[0017] For missing data, if the amount of missing data is small, the missing rate R is [missing information]. m For values ​​less than 3%, the mean-filled method is used; that is, for numerical data, the mean of the variable is used. Fill in missing values, the formula is: Where n is the number of non-missing data for this variable, x i Values ​​for non-missing data;

[0018] For categorical data, use the mode imputation method, which selects the category that appears most frequently as the imputation value for missing values;

[0019] The data standardization methods are as follows:

[0020] The Z-score standardization method is used, and the formula is as follows: Where x i The original data is represented by μ, which is the mean of the variable, and σ is the standard deviation of the variable. The standardized data has a mean of 0 and a standard deviation of 1.

[0021] In one embodiment, the specific method for constructing the key indicator system in step two is as follows:

[0022] Coverage performance metrics:

[0023] Impressions E: refers to the total number of times the ad is seen by users;

[0024] Reacher R: refers to the number of unique users who see the ad;

[0025] Audience Matching Score M: Measures the degree of match between the actual reachers and the target audience. The calculation formula is... Where N m To reach the number of people who match the characteristics of the target group, N rTo reach the total number of people;

[0026] Interaction effect index:

[0027] Click rate CTR: the calculation formula is Where C is the number of clicks, E is the exposure;

[0028] Average stay time Ts: refers to the average stay time of users after clicking on the ad page, the calculation formula is Where T si is the stay time of the i-th user, n is the number of users who click on the ad;

[0029] Interaction rate IR: including the proportion of user interaction behaviors such as comments, sharing, collecting, etc. The calculation formula is Where I is the total number of interactions, R is the number of people reached;

[0030] Conversion effect index:

[0031] Conversion rate CVR: calculated according to different conversion targets, the calculation formula is Where Con is the number of conversions, C is the number of clicks;

[0032] Cost per unit P c : refers to the average purchase amount of each customer, the calculation formula is Where T p is the total sales, N c is the number of customers who purchase;

[0033] Repeat purchase rate R p : refers to the proportion of customers who have purchased the product again, the calculation formula is Where N rep is the number of repeat customers, N c is the number of first-time purchase customers;

[0034] Cost effect index:

[0035] Cost per unit exposure CPM: the calculation formula is Where C t is the total cost of ad placement, E is the exposure;

[0036] Cost per click CPC: the calculation formula is Where C t is the total cost of ad placement, C is the number of clicks;

[0037] Cost per conversion CPA: the calculation formula is Where C t is the total cost of ad placement, Con is the number of conversions;

[0038] Return on Investment (ROI): the calculation formula is Where T p is the total sales, C t is the total cost of advertising.

[0039] In one of the embodiments, in step three, the specific method for determining the index weight by using the analytic hierarchy process is as follows:

[0040] Establish a hierarchical model:

[0041] The target of the advertising effect analysis is taken as the top layer, the four dimensions of coverage effect, interaction effect, conversion effect, and cost effect are taken as the middle layer, and the specific indexes under each dimension are taken as the bottom layer.

[0042] Construct a judgment matrix:

[0043] The importance of the indexes in the same level relative to the indexes in the previous level is compared two by two, the 1-9 scale method is used, 1 represents that the two indexes are equally important, and 9 represents that one index is extremely important than the other index, and the judgment matrix A=(a ij ) n×n is constructed, where a ij represents the importance degree of the ith index relative to the jth index, and a ii =1;

[0044] Calculate the weight vector:

[0045] Calculate the maximum eigenvalue λ max of the judgment matrix and the corresponding eigenvector W, normalize the eigenvector to obtain the weight w i of each index, which satisfies

[0046] Consistency check:

[0047] In order to ensure the rationality of the judgment matrix, consistency check is needed; calculate the consistency index where n is the order of the judgment matrix; then calculate the consistency ratio CR according to the average random consistency index RI. When CR<0.1, it is considered that the judgment matrix has satisfactory consistency, and the weight distribution is reasonable; otherwise, the judgment matrix needs to be reconstructed.

[0048] In one of the embodiments, in step four, the specific method for constructing the dynamic effect evaluation model is as follows:

[0049] Time series decomposition:

[0050] The time series data of the advertising effect index is decomposed into a trend item T t , a periodic item St Seasonal items Se t and random item R t That is, Y t =T t +S t +Se t +R t , where Y t Let be the index value at time t; the trend term and period term are extracted using the moving average method, the seasonal term is extracted using the seasonal index method, and the remaining part is the random term;

[0051] Dynamic weight adjustment:

[0052] The weights of the metrics are dynamically adjusted according to different stages of the advertising campaign. Let the initial weight be w. i0 The adjustment factor is α t α t The weight w at time t is determined based on the characteristics of the delivery phase. it =w i0 ×α t And satisfy

[0053] Dynamic composite score calculation:

[0054] By combining dynamic weights and standardized values ​​of each indicator, the dynamic comprehensive score of advertising placement at time t is calculated. Where x′ it Let be the standardized value of the i-th index at time t.

[0055] In one embodiment, step 5 involves the following specific analysis of influencing factors:

[0056] Variable selection:

[0057] The dynamic comprehensive score S of advertising performance t As the dependent variable, the factors that affect the advertising effect are selected as independent variables, including the placement channel X1, placement time X2, advertising content X3, and market competition level X4.

[0058] Multiple regression analysis:

[0059] Analyze the various factors that affect the effectiveness of advertising, identify the key influencing factors, and provide a basis for adjusting advertising strategies;

[0060] Establish a multiple linear regression model: S t =β0+β1X 1t +β2X 2t +…+β k X kt +ε t Where β0 is a constant term, βi is the regression coefficient, reflecting the influence degree of independent variable X i on dependent variable S t ; ε t is the random error term.

[0061] The regression coefficient is estimated by least square method, and the calculation formula is where X is the independent variable matrix, Y is the dependent variable vector, is the estimated value of the regression coefficient.

[0062] Regression model test:

[0063] Degree of fitting test: calculate the determination coefficient R where SSR is the regression sum of squares, SSE is the residual sum of squares, and SST is the total deviation sum of squares.

[0064] Significance test: including t-test of regression coefficient and F-test of regression equation.

[0065] T-test of regression coefficient is used to test whether each regression coefficient is significantly not 0, and the calculation formula is where is the standard error of the estimated value of the regression coefficient.

[0066] F-test of regression equation is used to test whether the regression equation as a whole is significant, and the calculation formula is where k is the number of independent variables, and n is the sample size.

[0067] Key influencing factor identification:

[0068] According to the size of the regression coefficient and the significance test result, the key factors that have significant influence on the advertising effect are identified.

[0069] In one of the embodiments, step six: effect prediction and optimization suggestion is as follows:

[0070] Effect prediction:

[0071] The ARIMA model time series prediction method is used to predict the dynamic comprehensive score of advertising effect.

[0072] The form of ARIMA model is ARIMA (p, d, q), where p is the number of autoregression terms, d is the number of differences, and q is the number of moving average terms. First, the stationarity of the dynamic comprehensive score sequence is tested. If the sequence is not stationary, it is made stationary by d times difference. Then the values of p and q are determined, the ARIMA model is established and the parameters are estimated, and finally the model is used for prediction to get the predicted value of advertising effect in the future period

[0073] Optimization suggestion generation:

[0074] According to the key influence factor analysis result and the effect prediction situation, specific optimization suggestions are made.

[0075] In one of the embodiments, step seven: effect tracking and feedback adjustment, the effect of advertising is tracked, and the analysis method and advertising strategy are adjusted in time according to the feedback result;

[0076] Effect tracking mechanism is established:

[0077] Set a regular effect tracking cycle, collect the latest advertising data and effect indicators, calculate the dynamic comprehensive score, and compare it with the predicted value; calculate the prediction error Among them is the predicted value, S t is the actual value; calculate the mean absolute error and the root mean square error are used to measure the accuracy of the prediction;

[0078] Feedback adjustment:

[0079] If the prediction error is large, analyze the causes of the error.

[0080] The beneficial effects of the present application are: compared with the prior art, in the present application, the single index limitation is broken through, the correlation is analyzed through multi-index linkage, and the one-sided cognition is avoided; with the help of time series decomposition and dynamic weight adjustment, the dynamic change of the investment is dealt with; through data preprocessing to ensure data quality, help enterprises to accurately and comprehensively analyze the advertising effect, and improve the investment efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0081] Figure 1 is the flowchart of the present application; DETAILED DESCRIPTION

[0082] In the description of the present application, it should be noted that the terms "positive", "upper", "lower", "left", "right", "vertical", "horizontal" and the like indicate the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore it cannot be understood as a limitation on the present application.

[0083] Reference Figure 1 In order to solve the problems in the background art, the present application proposes the following technical scheme: 1. An advertising effect analysis method, specifically comprising the following methods:

[0084] Step one: data collection and preprocessing, clearly need to collect the type of data, including advertising data (such as the channel, time, amount of money, etc.), user behavior data (such as the number of views, clicks, dwell time, purchase behavior, etc.) and market environment data (such as industry competition, holiday factors, economic situation, etc.); Set data collection period, according to the length and characteristics of the advertising, determine to carry out data collection in day, week, month, etc. To ensure the continuity and timeliness of the data; In the preprocessing of the collected data;

[0085] In step one, data preprocessing includes data cleaning and data standardization;

[0086] The data cleaning method is as follows:

[0087] Remove duplicate data: through data comparison, find out the duplicate records and delete them, avoid repeated data interference on the analysis results.

[0088] Let the number of duplicate data be D r , the total amount of original data is D t , then the proportion of duplicate data is

[0089] When Rr exceeds the set threshold (such as 5%), you need to check the problem in the data collection process;

[0090] For missing data, different processing methods are used according to the specific situation. If the amount of missing data is small, the missing rate R m <3%, the mean filling method is used, that is, for numerical data, the mean value of the variable is used to fill in the missing value, the formula is Where n is the number of non-missing data of the variable, x i is the value of non-missing data;

[0091] For categorical data, the mode filling method is used, that is, the category with the most occurrences is selected as the filling value of the missing value. If the amount of missing data is large, the missing reason needs to be analyzed, and if necessary, the data is re-collected or the sample where the data is located is removed. Correct abnormal data: identify abnormal data by drawing box plot, scatter plot and other methods. For abnormal data, first verify its authenticity, if it is data entry error, correct it; If it is a real extreme value, according to its influence on the analysis results, decide whether to keep or properly handle (such as using the truncation method, replacing the value beyond a certain range with the boundary value of the range).

[0092] The data standardization method is as follows:

[0093] ​In order to eliminate the dimensional influence between different data indicators and facilitate subsequent comprehensive analysis, the data needs to be standardized. The Z-score standardization method is adopted, and the formula is Wherein x i is the original data, μ is the mean of the variable, σ is the standard deviation of the variable, the mean of the standardized data is 0, and the standard deviation is 1, so that the data of different indicators are comparable.

[0094] The above technical solutions are explained as follows: by clearly defining the multi-dimensional data collection range, the data comprehensiveness is ensured, and key information such as advertising delivery, user behavior and market environment is covered. The data cleaning link effectively removes duplicate data, processes missing values and corrects abnormal data, ensuring data quality and avoiding interference with the analysis results. Z-score standardization eliminates dimensional influence, making different indicators comparable. Its beneficial effects are to provide a high-quality, standardized data basis for subsequent analysis, reduce data bias, make subsequent index calculation and model building more reliable, and solve the problems of low data quality and poor usability in the prior art, laying a solid data foundation for accurate analysis.

[0095] Step two: build a key indicator system, thereby building an indicator system from four dimensions of advertising delivery coverage effect, interaction effect, conversion effect and cost effect.

[0096] In step two, the specific method for building the key indicator system is as follows:

[0097] Coverage effect indicators:

[0098] Exposure E: refers to the total number of times the advertisement is seen by users.

[0099] Reach R: refers to the number of independent users who see the advertisement.

[0100] Coverage population matching degree M: measures the matching degree of actual reach and target population, and the calculation formula is Wherein N m is the number of people in the reach who meet the characteristics of the target population, and N r is the total number of people reached.

[0101] Interaction effect indicators:

[0102] Click-through rate CTR: the calculation formula is Wherein C is the number of clicks on the advertisement, and E is the exposure. This indicator reflects the degree of user interest in the advertisement.

[0103] Average dwell time Ts: refers to the average dwell time of users on the advertisement page after clicking on the advertisement, and the calculation formula is Wherein T si is the dwell time of the i-th user, and n is the number of users who click on the advertisement.

[0104] Interaction rate IR: the proportion of interactive behaviors such as comments, sharing, and collection of users on the advertisement, the calculation formula is Wherein I is the total number of interactive behaviors, and R is the reached person.

[0105] Conversion effect index:

[0106] Conversion rate CVR: calculated according to different conversion targets (such as registration, purchase, etc.), the calculation formula is Wherein Con is the number of completed conversions, and C is the number of clicks.

[0107] Purchasing single price P c : refers to the average purchase amount of each customer, the calculation formula is Wherein T p is the total sales, and N c is the number of purchasing customers. p : refers to the proportion of customers who have purchased the product again, the calculation formula is Wherein N rep is the number of repeat purchasing customers, and N c is the number of first-time purchasing customers.

[0108] Cost effect index:

[0109] Cost per thousand impressions CPM: the calculation formula is Wherein C t is the total cost of advertising, and E is the number of exposures. This index represents the cost required to obtain 1000 exposures.

[0110] Cost per click CPC: the calculation formula is Wherein C t is the total cost of advertising, and C is the number of clicks.

[0111] Cost per conversion CPA: the calculation formula is Wherein C t is the total cost of advertising, and Con is the number of conversions.

[0112] Return on investment ROI: the calculation formula is Wherein T p is the total sales, and C t is the total cost of advertising. This index reflects the profit level of advertising.

[0113] The technical solutions are explained as follows: an index system is constructed from four dimensions of coverage, interaction, conversion and cost, breaking through the limitation of single index. The coverage effect index reflects the reach and precision of advertisement, the interaction effect index reflects user interest, the conversion effect index measures the ultimate value realization, and the cost effect index evaluates the input-output ratio. The formulas of each index are clear and the logic is clear, such as CTR reflecting user click willingness and ROI reflecting profit level. The beneficial effect is to comprehensively cover the effect of each link of advertisement delivery, realize multi-dimensional and stereoscopic evaluation, avoid one-sidedness, help enterprises grasp the advertisement performance from the overall perspective, and solve the problem of incomplete analysis caused by single index in the prior art.

[0114] Step three: index weight determination, adopting the analytic hierarchy process to determine the index weight

[0115] Since different indexes have different importance in the analysis of advertisement delivery effect, each index needs to be given a corresponding weight. The method.

[0116] In step three, the specific method of adopting the analytic hierarchy process to determine the index weight is as follows:

[0117] Establish a hierarchical structure model:

[0118] The target of advertisement delivery effect analysis is taken as the top layer, the four dimensions of coverage effect, interaction effect, conversion effect and cost effect are taken as the middle layer, and the specific indexes under each dimension are taken as the bottom layer.

[0119] Construct a judgment matrix:

[0120] The importance of indexes in the same layer relative to the indexes in the previous layer is compared two by two, 1-9 scaling method is adopted, 1 indicates that the two indexes are equally important, and 9 indicates that one index is extremely important than the other index, and a judgment matrix A=(a ij ) n×n is constructed, where a ij indicates the importance degree of the i-th index relative to the j-th index, and a ii =1.

[0121] Calculate the weight vector:

[0122] Calculate the maximum eigenvalue λ max of the judgment matrix and the corresponding eigenvector W, normalize the eigenvector to obtain the weight w i of each index, and satisfy

[0123] Consistency check:

[0124] In order to ensure the rationality of the judgment matrix, consistency check is needed. Calculate the consistency index Where n is the order of the judgment matrix. Then, the consistency ratio is calculated based on the average random consistency index RI (which can be obtained by looking up a table). When CR < 0.1, the judgment matrix is ​​considered to have satisfactory consistency and the weight allocation is reasonable; otherwise, the judgment matrix needs to be reconstructed.

[0125] The above technical solution is explained as follows: The Analytic Hierarchy Process (AHP) is used to determine weights. By establishing a hierarchical structure, constructing a judgment matrix, calculating weight vectors, and performing consistency checks, the importance of each indicator is scientifically allocated. Compared to subjective assignment, this method combines expert experience with mathematical logic to ensure the rationality of weights. Its beneficial effects include reflecting the differentiated importance of different indicators in the analysis, avoiding distortion of results caused by treating indicators equally, making the comprehensive evaluation more aligned with actual business needs, and solving the problems of arbitrary weight determination and lack of scientific basis in existing technologies, thereby improving the scientific rigor and credibility of the analysis results.

[0126] Step 4: Construction of a dynamic effect evaluation model;

[0127] In step four, the specific method for constructing the dynamic effect evaluation model is as follows:

[0128] Time series decomposition:

[0129] Decompose the time series data of advertising performance metrics (such as click-through rate, conversion rate, etc.) into a trend term (Tt) and a period term (S). t ), seasonal items (Se) t ) and random terms (R) t ), that is, Y t =T t +S t +Se t +R t , where Y t Let be the index value at time t. The trend and periodic terms are extracted using the moving average method, the seasonal term is extracted using the seasonal index method, and the remaining part is the random term.

[0130] Dynamic weight adjustment:

[0131] The weights of the metrics are dynamically adjusted according to different stages of advertising campaigns (e.g., initial, mid-term, and late-term). Let the initial weight be w. i0 The adjustment factor is α t (α t The weight of the target audience (w) is determined based on the characteristics of the campaign phase (e.g., increasing the weight of coverage metrics in the early stages and increasing the weight of conversion metrics in the later stages). it =w i0 ×α t And satisfy

[0132] Dynamic comprehensive score calculation:

[0133] The dynamic comprehensive score of the advertisement at the t time is calculated by combining the dynamic weight and the standardized value of each index Where x′ it is the standardized value of the i-th index at the t time. Through the change of the dynamic comprehensive score, the dynamic change of the advertisement effect can be intuitively understood.

[0134] The advertisement effect will change over time, so it is necessary to build a dynamic effect evaluation model to reflect the dynamic change trend of the advertisement effect.

[0135] The above technical solutions are explained as follows: Through time series decomposition to separate trend, cycle, season and random factors, the dynamic change law of the index is revealed. The dynamic weight adjustment optimizes the weight distribution according to the characteristics of the delivery stage, and the dynamic comprehensive score intuitively reflects the effect change. The beneficial effect is to fully consider the timeliness and dynamics of the advertisement effect, overcome the limitations of static analysis, track the effect fluctuation in real time, help enterprises find problems in the delivery process in time, solve the problem of ignoring time factors and lagging analysis results in the prior art, and improve the sensitivity and control of the change of the advertisement effect.

[0136] Step five: influence factor analysis;

[0137] In step 5, the influence factor analysis is as follows:

[0138] Variable selection:

[0139] The dynamic comprehensive score S t of the advertisement effect is taken as the dependent variable, and factors that may affect the advertisement effect are selected as the independent variables, such as delivery channel (X1), delivery time (X2), advertisement content (X3), market competition degree (X4), etc.

[0140] Multiple regression analysis

[0141] Various factors affecting the advertisement effect are analyzed to find out the key influencing factors and provide a basis for adjusting the advertisement delivery strategy. A multiple linear regression model is established: S t = β0+ β1X 1t + β2X 2t + … + β k X kt + ε t , where β0 is a constant term, β i is a regression coefficient, reflecting the influence of the independent variable X i on the dependent variable S t , and ε tis the random error term. The regression coefficients are estimated by least squares method, and the calculation formula is where X is the independent variable matrix, Y is the dependent variable vector, is the estimated value of the regression coefficient.

[0142] Regression model test:

[0143] Perform goodness-of-fit test: calculate the coefficient of determination where SSR is the regression sum of squares, SSE is the residual sum of squares, and SST is the total deviation sum of squares. R 2 The closer to 1, the better the regression model fits the data. Perform significance test: including the t-test of regression coefficient and the F-test of regression equation. The t-test is used to test whether each regression coefficient is significantly not 0, and the calculation formula is where is the standard error of the estimated value of the regression coefficient. The F-test is used to test whether the regression equation as a whole is significant, and the calculation formula is where k is the number of independent variables, and n is the sample size.

[0144] Key influencing factor identification:

[0145] According to the size of the regression coefficient and the results of the significance test, identify the key factors that have a significant impact on the advertising effect. The larger the absolute value of the regression coefficient, the greater the impact of the factor on the advertising effect; if the regression coefficient is positive, it means that the factor is positively correlated with the advertising effect, and vice versa.

[0146] The above technical solutions are explained as follows: taking the dynamic comprehensive score as the dependent variable, selecting multiple independent variables for multiple regression analysis, and identifying key influencing factors through model testing. Regression quantifies the degree of influence of factors, and significance test ensures the reliability of the results. The beneficial effect is the core factor of accurately positioning the impact of advertising effect, and the direction and strength of each factor are determined to provide targeted basis for strategy adjustment, solving the problem of difficult to determine the effect driving factor in the prior art, and enabling enterprises to focus on key links to optimize the placement strategy.

[0147] Step six: effect prediction and optimization suggestion;

[0148] In step six: effect prediction and optimization suggestion, the specific steps are as follows:

[0149] Effect prediction:

[0150] The time series prediction method (such as ARIMA model) is used to predict the dynamic comprehensive score of the advertising effect. The general form of the ARIMA model is ARIMA (p, d, q), where p is the number of autoregressive terms, d is the difference frequency, and q is the number of moving average terms. First, the stationary test (such as ADF test) is performed on the dynamic comprehensive score sequence. If the sequence is not stationary, it is made stationary by d times difference. Then the values of p and q are determined (by observing the autocorrelation function ACF and partial autocorrelation function PACF), the ARIMA model is established and the parameters are estimated, and finally the model is used for prediction to obtain the prediction value of the advertising effect in the future period (m is the prediction step number).

[0151] Optimization suggestion generation

[0152] According to the key influencing factor analysis result and the effect prediction situation, specific optimization suggestions are made. For example, if the advertising channel is a key influencing factor and the effect of a certain channel is better, it is recommended to increase the proportion of advertising in that channel; if the advertising content has a significant impact on the effect, it is recommended to optimize the advertising content to improve the user's attention and interactivity. At the same time, combined with the cost-effectiveness index, the expected cost and expected revenue of different optimization schemes are calculated, and the scheme with the highest cost-effectiveness ratio is selected. Let the expected cost of the optimization scheme k be C k , the expected revenue be R k , and the cost-effectiveness ratio be The scheme with the maximum Bk is selected as the optimal scheme.

[0153] The above technical solutions are explained as follows: the ARIMA model is used to predict the effect trend, the cost-effectiveness ratio is used to select the optimal scheme, and specific optimization suggestions are generated. The prediction result provides guidance for future advertising, and the optimization suggestion takes into account the effect and cost. The beneficial effect is to realize the forward-looking grasp of the advertising effect, help enterprises plan strategies in advance, and at the same time realize the optimal allocation of resources through cost-effectiveness analysis, solving the problems of lack of prediction ability and insufficient scientificity of optimization scheme in the prior art, improving the initiative and economy of advertising.

[0154] Step seven: effect tracking and feedback adjustment, tracking the advertising effect, and adjusting the analysis method and advertising strategy in a timely manner according to the feedback result.

[0155] In step seven: effect tracking and feedback adjustment, tracking the advertising effect, and adjusting the analysis method and advertising strategy in a timely manner according to the feedback result. The specific method is as follows.

[0156] Effect tracking mechanism

[0157] Set a regular effect tracking cycle (such as daily, weekly), collect the latest advertising data and effect indicators, calculate the dynamic comprehensive score, and compare it with the predicted value. Calculate the prediction error wherein is the predicted value, S t is the actual value.

[0158] Calculate the mean absolute error

[0159] and the root mean square error for measuring the accuracy of the prediction.

[0160] Feedback adjustment

[0161] If the prediction error is large (such as MAE or RMSE exceeds the set threshold), analyze the cause of the error, which may be that the model parameters need to be adjusted, the influencing factors have changed, etc. According to the analysis result, re-adjust the parameters of the dynamic effect evaluation model, the influencing factor analysis model or the prediction model. According to the effect tracking result, if the advertising effect does not reach the expected target, adjust the advertising strategy in time, such as changing the delivery channel, adjusting the delivery time, optimizing the advertising content, etc., and re-perform the effect analysis and prediction, forming a closed-loop optimization process.

[0162] The above technical solutions are explained as follows: a regular tracking mechanism is established, the prediction accuracy is evaluated through the prediction error indicator, the model parameters and the delivery strategy are adjusted according to the error reason, and a closed-loop optimization is formed. The beneficial effect is to ensure that the analysis method and the delivery strategy continuously adapt to the actual situation, correct the deviation in time, maintain the timeliness and effectiveness of the effect analysis, and solve the problem of disconnection between analysis and practice, lack of dynamic adjustment mechanism in the prior art, and realize the continuous improvement of the advertising effect.

[0163] Although embodiments of the present application have been shown and described, the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing the effectiveness of advertising campaigns, characterized in that, Specifically, the following methods are included: Step 1: Data Collection and Preprocessing. Define the types of data to be collected, including advertising data, user behavior data, and market environment data; set the data collection cycle, determining whether to collect data on a daily, weekly, or monthly basis based on the duration and characteristics of the advertising campaign to ensure data continuity and timeliness; and preprocess the collected data. Step 2: Construct a key performance indicator (KPI) system, which will be built around four dimensions: coverage, interaction, conversion, and cost-effectiveness of advertising campaigns. Step 3: Determine the indicator weights using the Analytic Hierarchy Process (AHP). Step 4: Construction of a dynamic effect evaluation model; Step 5: Analysis of influencing factors; Step Six: Result Prediction and Optimization Suggestions; Step 7: Performance tracking and feedback adjustment. Track the performance of the advertising campaign and adjust the analysis methods and advertising strategies in a timely manner based on the feedback results.

2. The method for analyzing the effectiveness of advertising placement according to claim 1, characterized in that, In step one, data preprocessing includes data cleaning and data standardization; The data cleaning methods are as follows: Remove duplicate data: Identify and delete duplicate records by comparing data. Let the number of duplicate records be D. r The total amount of original data is D t The percentage of duplicate data When Rr exceeds the set threshold, it is necessary to check for problems in the data collection process; For missing data, if the amount of missing data is small, the missing rate R is [missing information]. m For values ​​less than 3%, the mean-filled method is used; that is, for numerical data, the mean of the variable is used. Fill in missing values, the formula is: Where n is the number of non-missing data for this variable, x i Values ​​for non-missing data; For categorical data, use the mode imputation method, which selects the category that appears most frequently as the imputation value for missing values; The data standardization methods are as follows: The Z-score standardization method is used, and the formula is as follows: Where x i The original data is represented by μ, which is the mean of the variable, and σ is the standard deviation of the variable. The standardized data has a mean of 0 and a standard deviation of 1.

3. The advertising effectiveness analysis method according to claim 2, characterized in that, In step two, the specific method for constructing the key indicator system is as follows: Coverage performance metrics: Impressions E: refers to the total number of times the ad is seen by users; Reacher R: refers to the number of unique users who see the ad; Audience Matching Score M: Measures the degree of match between the actual reachers and the target audience. The calculation formula is... Where N m To reach the number of people who match the characteristics of the target group, N r To reach the total number of people; Interaction effectiveness metrics: Click-through rate (CTR): Calculated using the following formula. Where C represents the number of clicks on the ad, and E represents the number of impressions; Average dwell time Ts: refers to the average time a user spends on the ad page after clicking on the ad. The calculation formula is as follows: Where T si Let n be the dwell time of the i-th user, and n be the number of users who clicked on the advertisement. Interaction Rate (IR): This includes the percentage of users who comment, share, or save ads. The formula is as follows: Where I represents the total number of interactive actions, and R represents the person reached; Conversion performance metrics: Conversion Rate (CVR): Calculated based on different conversion goals, using the following formula: Where Con represents the number of conversions completed, and C represents the number of clicks; Average order value P c This refers to the average purchase amount per customer, calculated using the following formula: Where T p For total sales, N c For the number of customers to purchase; Repurchase rate Rp: refers to the percentage of customers who have purchased a product before making another purchase. The formula is as follows: Where N rep To increase the number of repeat customers, N c Number of first-time customers; Cost-effectiveness metrics: Unit exposure cost (CPM): The calculation formula is as follows Where C t E represents the total cost of advertising, and E represents the number of impressions. Cost per click (CPC): The calculation formula is as follows Where C t C represents the total cost of advertising, and C represents the number of clicks. Unit conversion cost (CPA): The calculation formula is as follows Where C t Where is the total cost of advertising, and Con is the number of conversions. Return on Investment (ROI): The formula is as follows Where T p For total sales, C t This represents the total cost of advertising.

4. The advertising effectiveness analysis method according to claim 3, characterized in that, In step three, the specific method for determining the indicator weights using the analytic hierarchy process is as follows: Establish a hierarchical structure model: The goal of advertising performance analysis is taken as the top layer, the four dimensions of coverage effect, interaction effect, conversion effect and cost effect are taken as the middle layer, and the specific indicators under each dimension are taken as the bottom layer. Construct the judgment matrix: For indicators at the same level, pairwise comparisons are made regarding their importance relative to indicators at the next higher level. A 1-9 scale is used, where 1 indicates that both indicators are equally important, and 9 indicates that one indicator is extremely more important than the other. A judgment matrix A = (a ij ) n×n , where a ij This indicates the importance of the i-th indicator relative to the j-th indicator, and a ii =1; Calculate the weight vector: Calculate the largest eigenvalue λ of the judgment matrix max The corresponding feature vector W is used to normalize the feature vector, thus obtaining the weight w of each indicator. i ,satisfy Consistency check: To ensure the rationality of the judgment matrix, a consistency check is required; the consistency index is calculated. Where n is the order of the judgment matrix; then, the consistency ratio is calculated based on the average random consistency index RI. When CR < 0.1, the judgment matrix is ​​considered to have satisfactory consistency and the weight allocation is reasonable. Otherwise, the judgment matrix needs to be reconstructed.

5. The advertising effectiveness analysis method according to claim 4, characterized in that, In step four, the specific method for constructing the dynamic effect evaluation model is as follows: Time series decomposition: Decompose the time series data of advertising effectiveness metrics into a trend term T. t Periodic term S t Seasonal items Se t and random item R t That is, Y t =T t +S t +Se t +R t , where Y t Let be the index value at time t; the trend term and period term are extracted using the moving average method, the seasonal term is extracted using the seasonal index method, and the remaining part is the random term; Dynamic weight adjustment: The weights of the metrics are dynamically adjusted according to different stages of the advertising campaign. Let the initial weight be w. i0 The adjustment factor is α t α t The weight w at time t is determined based on the characteristics of the delivery phase. it =w i0 ×α t And satisfy Dynamic composite score calculation: By combining dynamic weights and standardized values ​​of each indicator, the dynamic comprehensive score of advertising placement at time t is calculated. Where x′ it Let be the standardized value of the i-th index at time t.

6. The method for analyzing the effectiveness of advertising placement according to claim 5, characterized in that, In step 5, the analysis of influencing factors is as follows: Variable selection: The dynamic comprehensive score S of advertising performance t As the dependent variable, the factors that affect the advertising effect are selected as independent variables, including the placement channel X1, placement time X2, advertising content X3, and market competition level X4. Multiple regression analysis: Analyze the various factors that affect the effectiveness of advertising, identify the key influencing factors, and provide a basis for adjusting advertising strategies; Establish a multiple linear regression model: S t =β0+β1X 1t +β2X 2t +…+β k X kt +ε t Where β0 is a constant term, β i The regression coefficients reflect the independent variable X. i For dependent variable S t The degree of influence, ε t This is the random error term; The regression coefficients are estimated using the least squares method, and the calculation formula is as follows: Where X is the matrix of independent variables and Y is the vector of dependent variables. These are estimates of the regression coefficients; Regression model testing: Perform a goodness-of-fit test: calculate the coefficient of determination. Where SSR is the regression sum of squares, SSE is the residual sum of squares, and SST is the total sum of squares; Perform significance tests, including t-tests for regression coefficients and F-tests for regression equations; The t-test for regression coefficients is used to test whether each regression coefficient is significantly different from zero. The calculation formula is as follows: in This represents the standard error of the regression coefficient estimates; The F-test for the regression equation is used to test whether the overall regression equation is significant. The calculation formula is as follows: Where k is the number of independent variables and n is the sample size; Key influencing factors identification: Based on the magnitude of the regression coefficients and the results of the significance test, key factors that have a significant impact on the effectiveness of advertising are identified.

7. The advertising effectiveness analysis method according to claim 6, characterized in that, in step six, the effectiveness prediction and optimization suggestions are as follows: Expected Outcomes: The ARIMA model time series forecasting method is used to predict the dynamic comprehensive score of advertising effectiveness; The ARIMA model takes the form ARIMA(p, d, q), where p is the number of autoregressive terms, d is the number of differencing terms, and q is the number of moving average terms. First, the stationarity of the dynamic composite score sequence is tested. If the sequence is not stationary, it is differxed d times to make it stationary. Then, the values ​​of p and q are determined, the ARIMA model is built, and the parameters are estimated. Finally, the model is used to make predictions, obtaining the predicted advertising effectiveness over a future period. Optimization suggestions generated: Based on the analysis results of key influencing factors and the predicted effects, specific optimization recommendations are formulated.

8. The method for analyzing the effectiveness of advertising placement according to claim 7, characterized in that, In step seven, performance tracking and feedback adjustment, the effectiveness of the advertising campaign is tracked, and the analysis methods and advertising strategies are adjusted in a timely manner based on the feedback results. The specific methods are as follows: Establishment of an effect tracking mechanism: Set regular performance tracking cycles, collect the latest advertising data and performance metrics, calculate a dynamic composite score, and compare it with the predicted value; calculate the prediction error. in S is the predicted value. t The actual value is used; the mean absolute error is calculated. and root mean square error Used to measure the accuracy of predictions; Feedback Adjustments: If the prediction error is large, analyze the reasons for the error.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the processor being configured to execute program instructions stored in the memory to implement the steps of the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program instructions that can be executed by a processor to implement the steps of the method as described in any one of claims 1-8.

Images